Speaker A
When we started out, we didn't seem like we were going to be successful at all.
Anthropic co-founder Tom Brown shares his journey from early startups to building Claude, lessons from GPT-3, and insights on LLM system design.
Key Takeaways
- Building a successful AI startup requires resilience and a strong entrepreneurial mindset.
- Early product vision can be unclear but evolves through iteration and learning.
- Open, transparent communication and culture are critical for scaling AI teams.
- Specialized hardware and infrastructure are key to efficient large language model training and deployment.
- User-centric design and feedback loops help create effective AI assistants like Claude.
What the video covers
- Tom Brown discusses the early uncertain days of Anthropic competing with well-funded AI labs like OpenAI.
- He reflects on his startup journey from MIT graduate to early employee at YC startups and eventually co-founding Anthropic.
- Tom shares the mindset shift from working in big tech to being a 'wolf' entrepreneur responsible for survival and innovation.
- He recounts experiences at startups like Linked, Grouper, and Solid Stage, highlighting challenges in defining product vision.
- Tom explains his transition into AI research and joining OpenAI despite initial doubts about his skills.
- The video covers the cultural and engineering practices at Anthropic, including transparent communication and scaling efforts.
- Tom highlights the development and user focus behind Claude, Anthropic’s AI assistant.
- He discusses technical topics like infrastructure scaling, chip specialization for training vs inference, and rapid iteration cycles.
- Insights are shared on AI safety, product-market fit, and the importance of community feedback in shaping AI tools.
- The conversation concludes with reflections on future AI system design and the evolving landscape of AI research and deployment.
Chapters
- 00:00Early Days and Challenges at Anthropic
- 02:00Startup Mindset: From Big Tech to Entrepreneurial 'Wolf'
- 03:40Founding Startups and Defining Product Vision
- 05:20Transition to AI Research and Joining OpenAI
- 07:30Building Engineering Culture and Scaling at Anthropic
- 10:00Developing Claude: User Focus and Product-Market Fit
- 13:00Technical Insights: Infrastructure, Chips, and Iteration
- 15:30Future of AI Systems and Closing Thoughts
Full Transcript — Download SRT & Markdown
Speaker A
OpenAI had a billion dollars and all of this star power, and we had seven co-founders trying to build something. We didn't know if we were necessarily going to make a product or what the products
Speaker A
would look like. One thing that's interesting to look at is just that humanity is on track for the largest infrastructure buildout of all time. Tell us about the early days of Anthropic. So you had a general idea of
Speaker A
this sort of long-term mission that you wanted to do to, you know, not destroy humanity, but like what did you actually work on for the first year? How did that converge on an actual product?
Speaker A
[Music] Welcome back to another episode of The Light Cone. Today we've got a real treat, co-founder of Anthropic, Tom Brown.
Speaker A
Excited to be here. So Tom, one of the things that a lot of the people watching would love to figure out is you got started in tech at the age of 21, fresh from MIT. How does someone go from that
Speaker A
in 2009 to literally co-founding something as important as Anthropic? Summer 2009, Linked Language. Two of my friends had started that out. I think they had seen one of our other friends, Kyle Vote, kind of do a YC company. And
Speaker A
so it was in the water that that's a thing that we could try to do. They started out. I was the first employee back then. Yeah. You guys let me join for all the dinners and stuff like that
Speaker A
too. I could have instead gone to like a big tech company or something like that.
Speaker A
And I think probably just as a software engineer, I might have learned more software engineering skills. But I think by being there with the other co-founders without anyone telling us what to do basically, we had to figure out how to live, how to like the
Speaker A
company would die by default. I think in school there was a lot of a feeling of more of people would give me tasks and I would do the tasks. It's kind of like a dog waiting for food to be
Speaker A
fed to them in their bowl or something like that. And I think for that company it was more like wolves and we have to hunt our real food; otherwise, our kids are going to starve or something like that. I
Speaker A
think that mindset, I think, has been the most valuable mindset shift that I've had for trying to do bigger, more exciting things.
Speaker A
Yeah. Big tech just teaches you to work at a big tech company, whereas it's much more fun to be a wolf.
Speaker A
Yeah. How did you go from working at a friend's startup to then you started your own one? So Linked was, we ran the company for a bit. I ended up going back to school afterwards and
Speaker A
then when I left school I went to this company Mopub, the mobile advertising thing, right?
Speaker A
Yeah. Yeah. I was the first engineer there. I was like, okay, I want to be a wolf but I was really bad at programming also. I was very, very struggling as a software engineer. I know I want to do
Speaker A
more but I don't know how to do it yet. And so I think that was kind of an experience getting to scale something.
Speaker A
Winter 2012, one of my friends who was my smartest friend from college pitched me on let's go and start a YC company.
Speaker A
We did at the time Solid Stage. This was before Docker existed. And so the idea was to try to make it easier to do DevOps, but Docker didn't exist. So it was going to be a more flexible Heroku, which basically meant a more complicated
Speaker A
Heroku. And so we, I remember, we interviewed with you guys. I think folks didn't really understand what we were trying to build. I think we didn't really understand what we were trying to build that much when you're trying to do something new.
Speaker A
That's actually sometimes common. Yeah. I think we were an outlier there because we did our interviews and then we got called back driving back to San Francisco and TLB had written on the board like an angry frowny face and what
Speaker A
are you actually going to build? And so he wanted us to explain that. I guess we explained it enough or he was just like these guys still don't know what they're doing but maybe they'll figure it out. Halfway through I kind of
Speaker A
felt I still didn't actually understand what we were going to build and how we would attach a mission to it that I wanted to work on for my whole life.
Speaker A
Yeah. And so I left PG actually introed me to Michael Waxman who was the Grouper founder.
Speaker A
Yeah. So, Grouper was a dating app only it was novel in that you had three guys and three girls. This was before AI in a lot of ways. So there was a team of people
Speaker A
who would manually link people up, right? And they'd meet up at a bar and shenanigans would ensue.
Speaker A
Yes. Reliably shenanigans. People didn't always have a great time. I think you went on a couple group.
Speaker A
Okay. The pitch for Grouper for me, for why I was excited for it was just I was an incredibly awkward kid. What I wanted to do was basically have a thing that lets awkward people like me
Speaker A
go out and talk to other people, for me to talk to girls and feel like I was safe doing it with my friends around and stuff like that. And so I think who were going to be our employees
Speaker A
was important. I did all of our engineering interviews. Who would take someone? The only person who went on more was Greg Brockman.
Speaker A
I think he had a phase where every single week he would go and post on Slack or HipChat at the time.
Speaker A
New York and he was hanging out at the Recurse Center during this period. I think, oh, I think he was at Stripe. Maybe for part of it he was at Recurse.
Speaker A
Yeah. But he also had a phase where at Stripe he would just post in their thing every week, "I'm going on Grouper," for like a whole year.
Speaker A
So I ended up being close with Greg, which ended up being a connection to OpenAI.
Speaker A
What was the journey like? Because you started as you just graduated from MIT CS. You were 21. You became first an early employee for all these YC startups. Then you started your company just a couple years later. And what was
Speaker A
the path for you to eventually become the co-founder of Anthropic? It was a long path but it's pretty impressive.
Speaker A
How did you get there? I mean, it sounds like getting in touch with Greg at that moment.
Speaker A
And then you were one of the first, you know, a couple dozen people to join OpenAI as a result.
Speaker A
Yeah. So, I left Grouper June 2014, and I joined OpenAI I think a year later. I tried to build up courage to make the switch to try to learn AI research. At the time I was like, "Okay, it seems like
Speaker A
sometime in our lifetimes we might end up making transformative AI. If we do, that would be the biggest thing. Maybe there's some way that I could help out, but also I got a B minus in linear algebra in college." And so it seemed
Speaker A
like at the time you needed to be just a top superstar in order to try to help out with that at all. And so I think I had a lot of uncertainty about whether I would be able to help. And
Speaker A
also I'd had some success with startups. And so a lot of me was just like rather than trying to retool at this, I could try to do another startup or something like that. I feel like in that
Speaker A
period, going to work on AI research, which is not seen as a serious, like not like a practically serious thing to do, and you're in a world where people try and build companies and do
Speaker A
these really practical things, like what did your friends say? Were they like, "Oh, that's really cool you're going to go work on AI stuff," or was it not really? I think my friends were like, "That sounds weird and bad," kind of like
Speaker A
it doesn't really seem like AI safety is a thing we should be worried about; overpopulation on Mars doesn't make any sense. And my friends were also just like, "I don't know if you're going to be good
Speaker A
at that." Tough. I think that.
Speaker A
And what were you specifically at this point? Like you're reading research papers like what it what does it look like?
Speaker A
Yeah. So, first I was just kind of hanging out. I built like an art car for Titanic 7 and stuff like that.
Speaker A
Oh, that was fun. Yeah. Yeah. So I I spent like a whole summer like 3 months after Grouper doing that cuz honestly I was I was like kind of burned out for Grouper where I know startups like the highs are high like
Speaker A
the lows are low and we weren't working at the end. Our business wasn't succeeding. Our revenue was going down but I my main job still was like recruiting engineers and so I had to like pitch them on this dream that I had
Speaker A
had but I like no longer really sounds like a death march. And so I was super burnt out and I was like, "Okay, Tom, like chill out, do some yoga, like do some CrossFit, like build an art car." And
Speaker A
what was the hindsight? Like, you know, hindsight's 2020. What's the retrospective on like Grouper obviously attracted all these really, really smart people. The graphs were up and to the right and then it flatlined and maybe started declining. What happened?
Speaker A
I think that when we started the competition was like, "Okay, Cupid. It was all web- based." All web- based. The main problem that I think we were solving was the it's hard to like go and put yourself out there
Speaker A
and go like talk to someone new and they might just be like I don't want to talk to you. You seem weird. And so we solved that by just blind matching. Tinder came out while we were doing grouper and
Speaker A
Tinder solved that same problem with both people have to show interest before you get matched. So there's also no worries about getting rejected. And I think that they just had better that was a better solution to that same problem.
Speaker A
So good work Tinder. Good work all the swipers. I think that that that solved the like mission that we were trying to solve better than we solved it.
Speaker A
And then yeah, like when did you get serious about AI and just how did you approach it?
Speaker A
Three months of like kind of playing and having fun and then I ran out of money also when I had like my personal runway.
Speaker A
I I ran out and so I was like, okay, I think that I'm going to need 6 months of stealth study to have a shot at getting a job. At that point it was Deep Mind or Google Brain were the two places to do
Speaker A
work there or MIRI. Merie was the third one that I was like looking at. So I was like if I want to help out with that those are the three places to look at. I don't have any of the skills yet. I need
Speaker A
six months of self-study to feel like I would not be a drag on them and like actually be helping instead.
Speaker A
Can you um maybe explain a bit what was that self study like? Because I'm sure there's a lot of software engineers right now in their 20s are looking to retool to become AI researchers. What was what was that six months like? Even
Speaker A
though as you said you had uh gotten a B minus in linear algebra which is like core might have been a C++. I'm not I should check.
Speaker A
I'm going to keep telling that's pretty impressive where where you got to. Yeah. Yeah. It turned out okay. First I did a contract actually with Twitch um and like earned like enough to have that six months of runway. So I did like
Speaker A
three-month contract with Twitch and then I made a a plan to self-study. I don't think it's the right plan now for people too at least 2015. What did it looked like? It was like take a Corsera course on machine learning, try to solve
Speaker A
some Kaggle projects, read linear algebra done right, and uh I had a statistics textbook. I think I had YC alumni credits and so I bought like a GPU and I would like SSH into the GPU to like work through my courses for it.
Speaker A
And this is right after Yeah, it was already after Alex Knack, right? This is after Alex Neck. Yeah. So I was mostly doing image image classification stuff that I was trying to learn was like the thing that all the courses
Speaker A
would teach you to do. How did you get the open AI job? Because you were one of the few engineers. It was mostly researchers and they had pretty stacked team of researchers.
Speaker A
I messaged Greg um as soon as OpenAI was announced and I was like I'd love to help out in some way. I got a B minus in my linear algebra but I know some engineering. I've done a bit of
Speaker A
distributed systems work. if you guys need help, I'm like happy to mop floors if if you guys need I want to help out, however. And I think Greg was like, yeah, I think there's like a posity of people who he said posity, too. I was
Speaker A
like, fancy word there. There's a posity of people who know both machine learning and distributed systems. So, like, yes, you should do that. I think he introduced me to Peter Aiel also to help me put together like a little course for
Speaker A
myself, too. And then I checked in on with him, I think every month or something. And then after a couple months he was like oh we actually have a project which is uh we need to put together we want to play a game like
Speaker A
play games can you help uh make Starcraft environment and so I joined to like help them with the Starcraft uh environment. So that that ended up I think getting my foot in the door. I I didn't do any machine learning work with
Speaker A
them for the first nine months that I was there basically. And what did OpenAI feel like at this point? Like had it raised much funding?
Speaker A
Did it have like an office? which is what will do it. Did it feel like a startup?
Speaker A
So it was in the chocolate on top of the dandelion chocolate factory. Um this is after Greg's apartment. That's the after Greg's apartment. Yeah. So like right after Greg's apartment in the chocolate factory when it kicked off, right? It was like a billion dollars of
Speaker A
committed funding from Elon. It felt like it was like very solid. The other interesting milestone for you was when you got to build a lot of the engineering around the training for GBT.
Speaker A
Yeah. For GP3 for and how how what was that? Because you got from GPT2 was in TPUs, right?
Speaker A
Yep. And the big breakthrough in GPT3 was like use more compute and using GPUs.
Speaker A
Yep. So I ended up working at OpenAI for a year, left, went to Google Brain for a year, came back, and then GPT3 was 2018 through 2019 was like building up to GP3, which exactly as you said was like
Speaker A
scaling things up. I think that like Daario had seen the big trend of scaling laws basically. You published a paper for that.
Speaker A
Yeah. Yeah. And that's like a pretty important paper that now has withtood the test of time and we're living now the dream of it.
Speaker A
Definitely like seeing that line of reliably you get more intelligence if you spend more compute with the right recipe was the main thing that was at least for me was like this is a thing that's like happening happening now cuz
Speaker A
you could look even at the time we weren't spending very much money on the on the training jobs at the time and you could see that there was scaling there and then also Danny Hernandez did a paper at the time that showed uh how
Speaker A
much cheaper algorithmic efficiency was making stuff over time too and like those two things stack together that was like, oh wow, we're going to get a lot more intelligence over the next few years.
Speaker A
So, it was noteworthy and surprising when you saw it. Yeah. And I I think the thing that seemed the weirdest to me is like I'm not a physicist, but like all these physicists were were doing this stuff.
Speaker A
The like original scaling laws paper just the like very straight line over like 12 orders of magnitude. I'm just like 12 orders of magnitude is like just like a stupidly large amount of I've like never seen anything go over 12
Speaker A
orders of magnitude. that convinced me to definitely pivot all of my work into scaling which I I hadn't been doing before.
Speaker A
Can I ask a like kind of lay person question? I mean is it fair to say that the scaling law might show up in all of these other domains then they're like are there like two five 100 10,000 domains where the
Speaker A
scaling law could hold that we're just not investing into? Yeah. So I think in physics scaling laws hold all over the place which I didn't know at the time but um within physics like there's a whole field called
Speaker A
phenomenology that basically looks at various aspects of the world and then does those types of fits and they they find these like power law distributions all over the all over the place. This was like I think the first one that I
Speaker A
had ever seen in a um like computer science adjacent thing which I I think was like interesting and surprising and and at the time it was people were mad about it. They actually were like you're throwing money at GPUs or just like
Speaker A
wasting money. This is very wasteful. Yeah. That was sort of People are still mad about that.
Speaker A
Yes. Different people now but still people mad about it. Yeah. Yeah. I guess. Yeah. The researchers were mad at that too where it's like it's it's not elegant. you're just like brute forcing it. The like jester cap like stack more layers like
Speaker A
which I think I think like anthropics slogan I think is like do the stupid thing that works. That was a thing where like this was very clearly the very stupid thing that that works.
Speaker A
Can you uh tell us then how you ended up collecting the last infinity stone with? Yeah, with Enthropic because there's very few people in the world that have basically worked at OpenAI, deep mind and anthropic and you were
Speaker A
part of the team that spun off from GPD3. Yeah. And then started Anthropic. So how was how was that jump?
Speaker A
There were two teams there. That was the safety or and the scaling or were the two orgs that reported into Daario and Daniela. I think we had just like worked together extremely well. One thing I think that was great both at OpenAI and
Speaker A
and at Anthropic was just like we had a culture where like everything is on Slack 100% of things on Slack. And within that all public channels, great communication. I think that that group also was the group that took the scaling
Speaker A
laws the most seriously where it was like okay like this actually is going to be transformative. there's going to be a handoff where like humanity will hand off control to transformative AI AI at some point and hopefully like they'll be
Speaker A
aligned with us and like that'll be a good transition that goes well but it might not be the stakes are incredibly high and so I think that group was very focused on like how do we make sure that
Speaker A
that's taken seriously enough and that like we've built an institution that can handle the weight of that that ended up being the core group that left to join Anthropic and I think I think it wasn't clear at all to me that like that was
Speaker A
the right thing for the world at the time. In hindsight now, it seems like that was a good choice. I think what was kind of cool then too is when we started out, we didn't seem like we were going
Speaker A
to be successful at all. OpenAI had a billion dollars and like all of these all of this star power and we had seven co-founders in COVID like trying to build something and we didn't know if we were necessarily going to make a product
Speaker A
or what the products would look like. And so I think that what was interesting from that too is that all of the initial people who joined were there for the mission too. They all could have worked somewhere else for more prestige, more
Speaker A
more more money. People would have known what they were doing etc. Well stayed at that opening eye basically.
Speaker A
Exactly. Yeah. That that exactly that's been an interesting thing then that I think has been like the key to like letting our culture or like let our org scale. We're like 2,000 people now but we still have a thing where it doesn't
Speaker A
seem like politics have creeped in. And I think a lot of that is like the first hundred people all were just there for the mission. So like if something starts to go wrong, they'll like raise their hand and be like, "It seems like this
Speaker A
person might not be acting for the for the mission." YC's next batch is now taking applications. Got a startup in you? Apply at y combinator.com/apply.
Speaker A
It's never too early and filling out the app will level up your idea. Okay, back to the video.
Speaker A
Hey, tell us about the early days of anthropic. So the the seven you broke off from open AI, you had a general idea of the sort of like long-term mission that you wanted to do to you know not destroy humanity but
Speaker A
like how did what did you actually work on for the first year? How did that converge on an actual product? So first year the main thing that I tried to do was just build the training infrastructure that we needed to train a
Speaker A
model and then get the compute that we needed to train the model. Those were like my two main projects. all the other things that you need to do when you're like starting up a company too. So like set up a Brex account and like I don't
Speaker A
know like all all of that all of that stuff. We started out with seven co-founders. Within like a few months I think like 25 folks from OpenAI um overall had joined. So we had like a pretty substantial team that like
Speaker A
already knew how to work together too. And so that helped us get up and running faster.
Speaker A
And at what point did you launch the first product and when did things begin to actually start working? So the first product that we launched was after chatgpt. We had like a maybe nine months before chat gpt. We had a slackbot
Speaker A
version of like claude one. Oh yeah, we had that in the YC uh slack actually.
Speaker A
Yeah. Yeah. Yeah. I remember like Tom Blfield adding all of you guys to it.
Speaker A
It was really cool. And then I think that at the time though we didn't know whether or not we wanted to launch it as a product. We didn't know if doing so would be good for the world at the time. I think we hadn't
Speaker A
really thought through our theory of impact that much for like how we actually will make stuff work well.
Speaker A
Plus, I think actually in hindsight like if we had tried to launch it, we like wouldn't have had the serving infrastructure to have done it. And I think because we weren't sure whether or not we wanted to, we like hesitated for
Speaker A
too long on building that infrastructure, which I think is learning for for me. I mean, at this time, ChatGpt had not launched yet. Chat GPD hadn't launched and so I guess we didn't know that it would be a big deal too.
Speaker A
This is around the pandemic 2022. This is summer 2022. Yep. And then chat GPD launched fall 2022 and then we relaunched our API after that and then claude AI after that also. I think it didn't seem like it was working
Speaker A
basically until Claude 35 and coding. I think like really really like through that whole time then until about a year ago it seemed like it wasn't clear that we were going to end up being like a successful company.
Speaker A
We actually saw that in the startups because we kind of get a bit of a vibe check in terms of what is the preferred model for startups. So all of 2023 open AI was the response. Then things started
Speaker A
to turn in 2024 is when uh we saw claw 3.5 and especially sonnet was starting to get a market share per se in the YC batches going from single digit to at some point like 20 and to 30% and especially for
Speaker A
coding y became the default choice which was very interesting. Can you tell us about how that emergent behavior and the spikiness on that particular skill must be 80% now or 90.
Speaker A
Yeah. for coding even more especially now clock code. What was that? Was that on purpose or just kind of happened?
Speaker A
I think that we invested more in trying to make the model really good at code because we wanted the model to be good at code was one thing and you did it.
Speaker A
Yeah. And then I think seeing seeing the reaction of everyone to it was like okay yeah like let's go much harder on that also.
Speaker A
And this is before 3.5 sonnet. you'd already invested enough in coding to realize that that was really promising and you decided to double down.
Speaker A
I think this really was like individuals within the org being like we want to do coding uh before 35 sonnet and then when we saw 35 sonnet's really good product market fit that was good signal to like go go for that
Speaker A
and do you guys know like the day that you guys launched 3.5 sonnet did you know that you had something really special and this was going to be the turning point for the company or were you as surprised as openi when they
Speaker A
launched chat GBT and it just like unexpectedly took off? Yeah, I I wish that I wish that we had like more foresight on that, but no, I think I think it was surprising for us too like how how big of a deal it was. And then I
Speaker A
think 37 sonnet also like surprised us by how much it unlocked like agentic coding. I think for for each of these things, yeah, we move quite fast in rolling them out and so we really um often don't know what the results are
Speaker A
going to be there. I think it's what made a lot of these coding agent startups work. I mean there's a crazy story of Replet winning going to 100 million in uh just 10 months right there's cursor of course a
Speaker A
story and all built on all these with with sonnet I think that all all of those things have been surprising to me and then also just like in my working with claude too like I think I continue to be surprised
Speaker A
by like the type of stuff that it can do and I I do think with each one there's like more stuff that kind of unlocks but one of my friends was telling me that she had some code that she uh some code
Speaker A
source tool that she wanted to modify, but she didn't have the source code for it. She had the compiled binary, and she's like, "Claude, can you can you decompile this?" Like, "Yeah, can can you disassemble the assembly?" And
Speaker A
Claude Claude chewed on it for 10 minutes and like made a C version of it.
Speaker A
And so then she had the thing to which is insane. And she's like, "Yeah, like if I spent 3 days on it, I probably could have gotten the hex tables and like wrote a little code to do, but like
Speaker A
it did the whole thing, made up variable names for them, etc." So I do think that like we keep getting surprised by stuff that model has memorized all the hex tables it can think through try to work through it. I think we're going to
Speaker A
continue to be surprised by that sort of stuff too. If you pull like the YC founders they prefer using anthropic models for coding by like a huge margin that's much larger than what you would predict if you just
Speaker A
looked at the benchmark results. Yeah. So there there seems to be some X factor that makes people really like these models for coding. Do you know what it is and is it intentional in some way or it just came out of the black box
Speaker A
somehow? I think that the benchmarks benchmarks are like easy to game where I think that all the other big labs I think have teams where they like their whole job with the team is to like make the benchmarks scores good and we don't have
Speaker A
such a team. And so I think that I think that that is probably the biggest factor.
Speaker A
You don't teach to the test. We don't teach to the test cuz I I do feel like if you start doing that then like it has weird bad incentives. Maybe we could like put that team under marketing or something like that and
Speaker A
then ignore all the benchmarks. But I think that that's one reasons why there's some train test mismatch there.
Speaker A
So the evaluations are more qualitative but internally or you have your internal We have internal benchmarks. Yeah. But we don't we don't publish them.
Speaker A
And is it the internal benchmarks that the teams are really focused on improving? That's right. Yeah. So we have internal benchmarks that the team focuses on and improving and then we also have a bunch of tasks like I think that accelerating
Speaker A
our own engineers is like a top top priority for us too and so we we do a ton of like dog fooding there to make sure that it's helping with our folks too. Going back to go Golden Golden Gate
Speaker A
Claude, there's a lot of sort of inter the interpretability seems like it's a big part of it. And then most people would say that, you know, Claude's personality just feels better. And then how do you sort of at once be very
Speaker A
quantitative, but then also, you know, build evals around personality. The evals for personality are kind of complicated, too, for like how do you tell if like Claude has like a good heart or something like that? It's like hard to hard to know. Um, but I do think
Speaker A
that that's like Amanda Ascll's team's mandate is I think she describes it as like being like a a good world traveler where like it can like Claude goes and talks with all sorts of people from different backgrounds and like each of
Speaker A
the people should come from come to that being like I I like feel good about like this conversation that I've had interp really I think is like a long-term bet right where it's like right now the models aren't that scary but at some
Speaker A
point they're going to be more scary and so I think the hope there is to have some ability to know what's actually going on under the hood when it becomes more intense. Then more recently, Claude Code's been a real success. Can you talk
Speaker A
us through like how did that project get started internally? And again, was it like a uh did you like know this time it was going to work or was it a surprise?
Speaker A
Claude Code was um an internal tool also. So like try to help out our our engineers within Anthropic that uh yeah, Boris um had like hacked together.
Speaker A
There's an internal anthropic engineer wanting to build it for themselves for internal for other internal engineers. Yeah. For him and other internal engineers. And then um I think yeah I think we definitely didn't know that it would be successful out there.
Speaker A
And I I think I think to to some degree like we really had fully just bet on the API before that with the intention being like there's like so many so many startups out there with so many good
Speaker A
ideas. Who are we to like figure out what the right product is to build on top of this stuff? Everyone out there is going to build better stuff than us. And so put all of our effort into just
Speaker A
making the best possible API. And I think that this surprised me as like okay like we actually were able to make something that like as a product was like better than the other products out on the market for this agentic use. I
Speaker A
have like some theory that like part of that came from like a mind shift of seeing Claude as like the user uh for this thing too. For like link that we were like trying to build things for teachers were like our users for for
Speaker A
grouper it was like single people in New York mostly I guess. Um, for this I think really the the like users are the developers but also I think the users is Claude. It's like give Claude the right tools that Claude can actually do that
Speaker A
effectively help Claude get the right context to work effectively. This team was like the most focused on Claude as like a user which I think makes sense that you guys would understand Claude the best. Yeah, I I do think that that's
Speaker A
a place where like startup founders though like can can do that too. And I think that that's that's probably a rich vein for people to like make tools that are better for models as users.
Speaker A
That's the perfect anthropomorphization of like the LLM itself. Like the agent is one of the stakeholders is one of the users that you would go after and try to like empower.
Speaker A
Yeah. Yeah. Totally. which actually makes a lot of sense why you guys actually got MCP to work to do tool calling because a bunch of other labs had tried to do something and the standard that stuck that really took off
Speaker A
was yours. Yeah, I think that that seems like a similar one too where it's like model model focused going back to claw code. So like success is really exciting. It's also scary for like cursor and other companies that
Speaker A
have built on top of the API like what's your advice to founders building products like how should they think about building on the API but also worrying about like anthropic or one of the labs building something better than
Speaker A
they can build. I think I was kind of surprised that claude code like we we did build a thing that was like uh like the best in the market there too. It's not super clear to me what the big advantage was for us
Speaker A
for Claude code besides more empathy for Claude or something. That's actually I think that's actually really interesting insight. Like it seems like the thing that Yeah. you were building for a specific user that you knew really well that other people wouldn't have thought
Speaker A
to build for versus like you had some like intrinsic technology advantage. Yeah. Like I think a startup could could have done that same thing too, right?
Speaker A
Yeah. I think we're the most like developer focused lab. I think we're the most like API focused lab too. So I think we want to make sure that we have the best platform for people to build stuff on
Speaker A
cuz this thing is growing so incredibly quickly. like we're not going to be the fastest at figuring out all the ways that we need to empower Claude to do the work that connects Claude to the entire human business that's like human human
Speaker A
world is all designed for humans but like we need to get the models to be able to be productive members of uh the economy.
Speaker A
Are there like ideas or areas you would love to see developers building in or like areas you don't you you think are like underappreciated right now? Yeah, Claude code is like how do you get Claude to be a useful pair programmer
Speaker A
kind of um or like junior engineer. You've got like a level two or three or something like that that you can work with or like very spiky because also it can do the like weird disassembly stuff that like a super high
Speaker A
level suite would struggle with. Less good at knowing what type of work to do.
Speaker A
Needs kind of a lot of handholding. Needs a lot of context from it. That's like one very particular subset of work that can be done. Uh if you look at like all the stuff that happens in businesses besides that, it's like a very tiny
Speaker A
fraction of like all the work that's done in businesses that like a smart person who knows how to code and like use lots of tools but doesn't have that much context yet uh would want to do. So I think I think finding ways to coach
Speaker A
Claude or uh co coach whatever model to like do useful tasks for businesses seems like there's just like a huge huge space there.
Speaker A
So Tom, a big part of your job is like owning all the compute infrastructure that makes anthropic work. Can you talk about like what what is the compute infrastructure behind this giant thing?
Speaker A
Now, one thing that's interesting to look at is just that humanity is on track for like the largest infrastructure buildout of all time.
Speaker A
Now, this is going to be larger than the Apollo project, larger than the Manhattan project.
Speaker A
It'll be bigger than both of them next year if it keeps on the current trajectory, which is like roughly 3x per year increase in spending on AGI compute, which is just bonkers. Yeah.
Speaker A
Like 3x per year is wild. I think it's going to keep up on the 3x per year trajectory. It's already locked in for that for for next year and then it's a little bit open for for 2027.
Speaker A
I mean anecdotally internal to YC uh we can't get enough you know credits across all of the top frontier models including Claude. So you got to help us out a little bit.
Speaker A
Yeah. We're just I mean everyone's bottlenecked literally every you know it's like give me more intelligence. I can't have enough.
Speaker A
Yeah. And I know you guys have been looking at more hardware startups also for like more accelerators. I think that we will see more accelerators coming online to 2027. That's a good a good space. Also like data center tech I
Speaker A
think is a big one. Where are the bottlenecks for you guys now? Is it like getting enough electricity, getting enough GPUs, getting construction permits, power, people are using jet engines to get power. That's nuts.
Speaker A
Overall for the buildout, I think power is going to be the biggest bottleneck, especially power in the US. Like we want to build in the US. That's one of our biggest policy goals is to like get the US to like build more data centers,
Speaker A
permit more data centers, make it easier to build. Is the answer renewables or is it uh nuclear?
Speaker A
I I definitely I feel like yes, yes, all all of those things. I wish I wish that nuclear was easier to build.
Speaker A
Anthropic is the only major lab that uses not just one kind of GPU, but the GPUs from three different manufacturers.
Speaker A
Can you talk about that and how how how that strategy has played out? Yeah. Yeah. So we use um GPUs, TPUs and tranium. Downside of doing that is that we split our performance engineering teams across all of those platforms
Speaker A
which is a ton of extra work. The positive thing is it gives us the flexibility to both one like soak up that extra capacity because there there just is more of those altogether than just one and then two is we can use the
Speaker A
like right chips for the right jobs where some chips will be better for inference, some chips will be better for training and we can match the the right chips to the right jobs. So yeah, I think that that's kind of the the
Speaker A
trade-off there. I guess one cool thing is just connecting the dots through your career and how all of this compounded because you you were the one engineer building that change of the architecture from TPUs to GPUs back at OpenAI that
Speaker A
got GPD3 to actually scale and now you're in charge of that at a much much bigger scale year years later. I don't know if that kind of connected dots for you. The big move from TPUs to GPUs at
Speaker A
OpenAI I think was partly driven just that PyTorch was a better software stack on top of them than TensorFlow on top of TPUs. And I think that that then unlocked fast iteration where like if you have like a good reliable software
Speaker A
stack then you can experiment quickly just like build a whole system that works. I think that that's the thing that we really strive for now at Anthropic too is a challenge of having many more platforms is that it's harder
Speaker A
to write all the good software. I think building the muscle of knowing how to build that software well so that all of the people who build on top of that low level can have a great experience with it is the most important thing there.
Speaker A
Do you have advice for um kind of like a younger Tom version of yourself who now you've seen and went through this crazy journey? If someone was you back in the 20s living today and they wanted to ride
Speaker A
and join the AI revolution, what would you say to them? And very specifically something we see from a lot of hear from a lot of college students at the moment is they uh they don't know what like if they should stay
Speaker A
in college like are there going to be jobs for them what like how is the world going to change and what should they do?
Speaker A
taking more risks I think is is wise and then also trying to work on stuff where your friends would be really excited and impressed if you did it or a more idealized version of yourself would be really like proud of yourself if you
Speaker A
succeeded at it I think is like probably the thing that I would I would try to tell a younger version of myself more intrinsic less exttrinsic like don't chase these other credentials and getting the degree or whatever you know
Speaker A
working at Fang like those are just irrelevant as of today. Yeah exactly. That's all we have time for today. We'll see you guys next time.
Speaker A
[Music]
Topics:AnthropicTom BrownClaude AIGPT-3Large Language ModelsAI startupAI system designOpenAIYC startupsAI research
